Hi all!
As far as I could find out gmond has a set of metrics built-in at
compile time (a rather convenient set most of the time). If more
information regarding the node or the software running on that node is
necessary then gmetric can be used to publish that metric.
Hoping that the previous paragraph affirmation is correct I would like
to ask around some question, though I will also provide some possible
answers/thoughts about them:
1. Doesn't using gmetric (forking a process) consume a bit too much of
the system resources?
2. I need to monitor a JVM profile. Has anyone tried something like
similar? Any thoughts or ideas on best way to achieve this with
Ganglia?
3. I also need to monitor some database services... thoughts and ideas
most welcome.
Now for my though on the subject.
Forking is always a intrusion on the system, even if linux is rather
efficient at doing it. If it's possible to avoid then I guess it's the
best.
To monitor the JVM I though that the best choice is possible to use the
sources for gmetric and build a JVMPI based Ganglia Metric Client. Has
anyone developed such a thing yet? Is there wide interest in such a
thing?
For the database services, well, more or less as with the JVM: a
expansion of the gmetric so that queries are sent to the DB server,
parsed and then announced to Ganglia. Anyone interested?
Thanks for your patient with a newbie with Ganglia and with a long
mail,
Miguel Tavares
--
"Until they become conscious they will never rebel,
and until after they have rebelled they cannot become conscious."
- George Orwell's 1984 -
